Mar 18 2019
A research team at Politecnico di Milano has created an electronic circuit that can solve a system of linear equations in a single operation in the timescale of few tens of nanoseconds. The performance of this new circuit is greater not only to the classical digital computers, but also to the ultramodern quantum computers: it will be soon possible to create a new generation of computing accelerators that will transform the technology of artificial intelligence.
Solving a unit of linear equations means discovering the unidentified vector x which fulfills the equation Ax = b, where A is a matrix of coefficients and b is an identified vector. To solve this problem, a conventional digital computer performs an algorithm which involves a number of operations, thus translating into substantial time and energy consumption.
The new circuit, which has been created within the framework of the ERC European project RESCUE (Resistive switch computing beyond CMOS), solves systems of linear equations (Ax=b) thanks to an advanced technique of in-memory computing, where the coefficients of matrix A are stored in an exclusive device known as memristor. The memristor is capable of storing analog values, and so a memristor matrix can physically map a coefficient matrix A inside the circuit, thus powerfully speeding up the computation.
The memristor array has been created at the Clean Room of the Center for Micro and Nano Fabrication Polifab of Politecnico di Milano. The memristor circuit has been tested and tried out on a broad set of algebraic problems, such as the grading of internet websites and the solution of complex differential equations, for example, the Schrödinger equation for the computation of the quantum wavefunction for an electron. All these issues are cracked in a single operation.
These results have been reported in the prestigious journal PNAS of the National Academy of Science of the USA.